非局部混合噪音滤波器MNF的一种快速算法

A Fast Algorithm to Nonlocal Mixed Noise Filter MNF

对非局部混合噪音滤波器MNF提出一种快速算法(Fast-MNF).该算法是对目前滤波效果较好的混合噪音滤波器MNF的一种改进,它对邻近窗口像素按其均值、梯度和方差的相似程度进行预分类,把均值比、梯度比和方差比均在给定阈值范围内的像素分为一类,让其参与滤波运算,否则被忽略,以此提高滤波速度.实验结果表明,Fast-MNF算法的滤波效果优于MNF,且运算速度提高20%以上.

A fast algorithm of mixed noise filter （Fast-MNF） is proposed, which is an improvement to MNF of better filtering effect at present. Neighboring pixels are classified, according to the mean value and gradient and vafiance similarity. The mean value and gradient and variance of the ratio within the range of a given threshold of pixels are divided into a class. Let them participate in the filtering operation, otherwise ignored, which can improve the filtering speed. The experimental results show that filtering effect of the Fast-MNF is better than the MNF and computing speed increases more than 20%.